Speaker
Wei Sun
Description
We provide a performance portable implementation of SU(N) lattice gauge theory (LGT) simulations using the Kokkos parallel programming model, enabling efficient execution across diverse architectures, including x86 CPUs, Arm CPUs, and GPUs. By leveraging Kokkos’s abstractions for parallel execution and memory management, we map the gauge field operations of SU(N) LGT onto heterogeneous hardware while maintaining a single codebase. Preliminary benchmarks demonstrate good performance on NVIDIA V100/A100 GPUs and ARM-based processors. This work establishes a template for high performance, cross platform LGT simulations.